Data representation for clinical data and metadata
WP1: Data representation for clinical data and metadata Inconsistent terminology creates barriers to identifying common clinical entities in disparate information systems and automatic linking integrating clinical and genomic data Remove this barrier through the establishment of a consensus regarding domain-specific vocabularies, ontologies and data items. WP1 acts as Rosetta Stone data can be mobilised
Trials of the future New data types - data integration data sharing
Translation of clinical data to genomics and bioinformatics research Molecular data Clinical data Ontology framework Integrative analysis Therapeutic systems
Data Integration Structured data acquisition Data sharing and re-use WP1 Essential knowledge engineering: Ontologies, common data elements and data objects Quality assurance Optimal data exploitation
Systems biomedicine: Integrating physiology and genomics Sylvia B. Nagl PhD Cancer Systems Biology & Biomedical Informatics Department of Oncology
Translating clinical trials to genomics research Fast accumulating molecular data from fundamental and translational cancer research, genomics, proteomics etc. Clinical trials increasingly incorporate detailed molecular analysis of tumour and serum samples Challenges: Elucidation of the relationships between molecular data and physiology (Ontological) integration of multi-scale data Translation of data into knowledge understanding of cell and tumour systems
Trials of the future
Neo-tAnGo Secondary endpoints Tumour tissue analysis Prognostic and predictive markers and profiles Molecular profiling using expression and DNA microarrays Immunohistochemistry (IHC) and in situ hybridisation Serum proteomics - protein profiles
Markers and systems clinical parameters cell and tumour physiology proteins gene expression genome
Pathway modelling clinical parameters cell and tumour physiology
Pathway modelling clinical parameters cell and tumour physiology SCIpath project www.scipath.org.uk
Pathway modelling Pathways located midway between the genome and the phenotype, and can be conceptualised as an extended genotype or elementary phenotype. Intrinsically offer an integration framework for modelling of biological processes, and the translation of clinical studies of novel treatments to the molecular and genomic level. A formal framework for the expression of complex biological knowledge, assumptions and hypotheses in a form amenable to logical analysis and quantitative testing. Computational analysis is increasingly necessary as the scope and depth of information and knowledge, with the accompanying uncertainty, surpass the analytical capabilities of the unaided human mind.
Pathway modelling http://sbw.kgi.edu/
Gene expression Gene-protein expression Co-expression patterns Cluster analysis Gene copy number Simulation
Bayesian network modelling nodes grade R Decision support systems size Calculate prognostic probabilities P(Recurrence X) Logic Rules Williams & Williamson, 2005: www.kent.ac.uk/secl/philosophy/jw/
Bayesian nets as pathway models Bayesian networks are increasingly important for integrating biological data and for inferring cellular networks and pathways. Needham et al., Nature Biotechnology 24, 51-53 (2006)
Multi-scale predictive profiles Vertical integration using Bayesian nets Probability Logic
Knowledge integration Development of Bayesian network integration methods e.g., integration of predictive models from separate clinical trials integration of predictive parameters from clinical and molecular levels evolvability, integration of new knowledge, new data and new data types (e.g., epigenomics) Collaboration with Williams (CR UK) and Williamson (U of Kent)
Integration with Cancergrid SCIpath and Bayesian tools will implement CG standards Tools can be integrated in CG-caBIG
Conclusions Clinical trials incorporating molecular profiling create the challenge and opportunity for new integrative computational approaches Pathway modelling effects of differential gene expression Bayesian networks integration of molecular and clinical parameters for prediction, probability + logic Integration of genome structure, molecular processes and physiology ( vertical integration, multi-scale profiles) Systems understanding s.nagl@ucl.ac.uk